Single Cell Spatial Transcriptomic Profiling Identifies a LINE1 Associated Disarrayed Immune Microenvironment in Hepatocellular Carcinoma

Abstract

Purpose: Hepatocellular carcinoma (HCC) is a lethal malignancy driven by complex interactions between cancer cells, immune cells, and additional stromal cells in the tumor microenvironment (TME). The LINE1 retrotransposon is a ubiquitous repeat RNA whose de-repression leads to significant cancer cell-intrinsic and TME changes that promote aggressive tumor characteristics. We leveraged single cell spatial transcriptomic profiling to characterize the relationship between LINE1 and differences in the heterogeneous HCC TME.

Experimental design: We applied our profiling methodology to a cohort of 23 tissue specimens collected from patients who had undergone liver resection or transplantation and validated it in a partially-overlapping similar cohort of 39 specimens using RNA in-situ hybridization (RNA-ISH).

Results: We found that LINE1-high tumors and LINE1-high single HCC cells exhibited a de-differentiated, stem-like, and inflammatory phenotype. Furthermore, within individual tumors, LINE1 high cancer cells associated spatially with one another and excluded the larger, organized immune cell conglomerates seen in LINE1 low tumors. Finally, we found that LINE1 RNA expression correlated with worse overall survival in the larger expanded retrospective cohort.

Conclusions: Our study is the first to show a clearly disorganized immune TME in HCC driven by LINE1 expression, and this observation correlated with poor survival for patients whose tumors expressed large amounts of the LINE1 repeat RNA. These results provide further evidence of how effective anti-tumor immune responses contribute to cures after definitive surgery and may lead to novel biomarkers or drug targets in HCC.

Keywords: LINE1; Spatial transcriptomics; hepatocellular carcinoma; immune microenvironment; repeat RNAs; tumor microenvironment.

Figure 1.. Overview of spatial molecular imaging (SMI) experiment utilizing HCC FFPE sections.

(A) Experimental workflow of spatial transcriptomic profiling experiment. (B) Serial section H&E staining (left) and corresponding original immunofluorescence staining (right) for core tissue samples used in the SMI experiment from TMA2. For TMA1, see Figure S3. White rectangles depict fields of view (FOVs), within which SMI was conducted. Green: pan-cytokeratin, red: CD45, yellow: CK8/18, cyan: CD298/B2M. (C) Immunofluorescent staining used for cell segmentation in a representative FOV: TMA2, FOV60. (D) Cellpose cell segmentation results (left) and representation of SMI transcript detection (right) used to identify tumor and immune cells in the same FOV as in 1C.

Figure 2.. Subcellular spatial molecular imaging (SMI) accurately identifies cell types in HCC tumor microenvironment.

(A) Schematic of cell-typing workflow, which included both manual annotation and supervised clustering, applied to TMA1. The same workflow was applied to TMA2 (see methods). (B) Final cell type annotations visualized on the UMAP embedding for the complete SMI dataset (left). Cells mapped back onto their physical locations in space, pseudo-colored by final cell type annotation for the same FOV in 1C (right). (C) Dot plot depicting expression of marker genes found algorithmically (see methods) for each of the identified cell type clusters. (D) Bar plot depicting the composition of each FOV in the complete SMI dataset in terms of major cell type categories.

Figure 3.. HCC expression of LINE1 is associated with a de-differentiated, stem-like state.

(A) Bar plot depicting the definition of LINE1-ORF1 high and low patient groupings in the SMI dataset. The dashed line represents the upper tercile for net tumor cell LINE1-ORF1 counts per million (CPM) across all patients. (B) Volcano plot summarizing differential expression analysis comparing tumor cells in LINE1-ORF1 high patients and LINE1-ORF1 low patients, as defined in 3A. P-values are FDR-adjusted. Significance threshold = 0.05. (C) Heatmap depicting average normalized expression of selected differentially expressed genes identified in 3B among tumor cells in each patient. Values have been row scaled for visualization. (D) Box plots depicting the definition of LINE1-ORF1 high and low tumor cell groupings within each patient in the SMI dataset. The black crossbars represent the upper tercile for normalized LINE1-ORF1 expression within each patient. (E) Volcano plot summarizing differential expression analysis comparing LINE1-ORF1 high tumor cells and LINE1-ORF1 low tumor cells in a patient-paired manner. P-values are FDR-adjusted. Significance threshold = 0.05. LINE1-ORF1 not shown. (F) Patient-paired differences in total normalized expression of selected differentially expressed genes identified in 3E between LINE1-ORF1 high and low tumor cells. The crossbars represent means. (G) Gene set enrichment analysis plots generated from the results depicted in 3E.

Figure 4.. High LINE1 expression within HCC tissue leads to a disorganized, sparse immune microenvironment.

(A) Patient-paired differences in LINE1-ORF1 high and low tumor cells’ colocalization patterns with themselves and one another. ER = enrichment ratio (see methods). P-values computed by paired Wilcoxon Rank Sum Test and FDR-adjusted. For other cell types, see Figure S6a. (B) Representative spatial maps for two FOVs depicting tumor niches comprised mainly of cells with the same LINE1-ORF1 group identity, as predicted by 4A. (C) Neighborhood enrichment analysis depicting pairwise colocalization for all cell type pairs in LINE1-ORF1 high patients (left) and LINE1-ORF1 low patients (right). (D) Randomization testing for difference in immune organization score between LINE1-ORF1 high and low patients. The red dashed line represents the observed difference which was compared to the shown empirical null distribution (see methods) to compute the reported p-value. (E) Representative immunofluorescent staining images and corresponding spatial maps for four FOVs illustrating differences in immune cell organization in LINE1-ORF1 high patients (left) and LINE1-ORF1 low patients (right).

Figure 5.. Validation of LINE1 expression by RNA-ISH.

(A) Representative RNA-ISH image showing LINE1 expression. (B) Jitter plot depicting the definition of LINE1 high and low patient groupings in the RNA-ISH dataset. The dashed line represents the upper tercile for LINE1 counts per μm² across all patients. The crossbars represent means. P-value computed by Wilcoxon Rank Sum Test. (C) Association of LINE1 RNA-ISH quantification with HERVH, HERVK, and HSATII RNA-ISH quantification. P-values computed by Wald test. (D) Kaplan-Meier overall survival analysis for LINE1 high patients versus LINE1 low patients (p = 0.01 by log-rank test).